Mining big spatial data

• المؤلف الرئيسي: Ng, Zhi Kai
• مؤلفون آخرون: Gao Cong
• التنسيق: Final Year Project
• اللغة: English
• منشور في: Nanyang Technological University 2023
• الموضوعات: Engineering::Computer science and engineering::Data::Data structures; Engineering::Computer science and engineering::Computing methodologies::Image processing and computer vision
• الوصول للمادة أونلاين: https://hdl.handle.net/10356/171918

The purpose of representing road network is to provide a comprehensive view of the real-world road connections and features for analysis purposes. However, limitations in data unavailability hinder the creation of road network representations to precisely align with actual road layouts. This study investigates the usage of images to construct a road network representation and subsequently improving the performance of several downstream tasks. Three distinct image encoder architectures are used to obtain the image embeddings. Furthermore, several enhancement techniques are applied to further boost the performance of the images. My findings showcase the performance comparison between images, enhanced images, and baseline methods. Specifically, images showed superior performance when compared with baseline methods. Moreover, enhancements contribute slightly to improving image performance. The study highlights the usefulness of images in constructing road network representations. By capturing the visual information from images, the study introduced a novel approach in representing road networks compared to traditional methods.